UniNE at CLEF 2017: TF-IDF and Deep-Learning for Author Profiling

This paper describes and evaluates a strategy for author profiling using TF-IDF and a Deep-Learning model based on Convolutional Neural Networks. We applied this strategy to the author profiling task of the PAN17 challenge and show that it can be applied to different languages (English, Spanish, Portuguese and Arabic). As features, we suggest using a simple cleaning method for both models, and for the Deep-Learning model, a matrix of 2-grams of letters with punctuation marks, beginning and ending 2-grams, as features. Applying this strategy, we determine that the TFIDF-based model is the best one for language variety classification and that the Deep-Learning model achieve the highest accuracy on gender classification. The evaluations are based on four tweet collections (PAN AUTHOR PROFILING task at CLEF 2017).